Differentially Private Feature Release for Wireless Sensing: Adaptive Privacy Budget Allocation on CSI Spectrograms

TL;DR

This paper introduces an adaptive privacy budget allocation method for differentially private CSI spectrogram feature release in wireless sensing, enhancing privacy protection while maintaining high utility in various sensing tasks.

Contribution

It proposes a novel adaptive privacy budget allocation mechanism tailored to CSI spectrograms, improving privacy-utility trade-offs in wireless sensing applications.

Findings

Adaptive allocation improves accuracy over uniform perturbation.

Method reduces identity and membership inference leakage.

Enhanced privacy-utility balance demonstrated across multiple sensing tasks.

Abstract

Wi-Fi/RF-based human sensing has achieved remarkable progress with deep learning, yet practical deployments increasingly require feature sharing for cloud analytics, collaborative training, or benchmark evaluation. Releasing intermediate representations such as CSI spectrograms can inadvertently expose sensitive information, including user identity, location, and membership, motivating formal privacy guarantees. In this paper, we study differentially private (DP) feature release for wireless sensing and propose an adaptive privacy budget allocation mechanism tailored to the highly non-uniform structure of CSI time-frequency representations. Our pipeline converts CSI to bounded spectrogram features, applies sensitivity control via clipping, estimates task-relevant importance over the time-frequency plane, and allocates a global privacy budget across spectrogram blocks before injecting calibrated Gaussian noise. Experiments on multi-user activity sensing (WiMANS), multi-person 3D pose estimation (Person-in-WiFi 3D), and respiration monitoring (Resp-CSI) show that adaptive allocation consistently improves the privacy-utility frontier over uniform perturbation under the same privacy budget. Our method yields higher accuracy and lower error while substantially reducing empirical leakage in identity and membership inference attacks.